Reservoir and production engineers tasked with modeling or managing large oil fields containing hundreds of wells are faced with the reality of only being able to physically evaluate and manage a few individual wells per day. Individual well management may include performing tests to measure the rate of oil, gas, and water coming out of an individual well (from below the surface) over a test period. Other tests may include tests for measuring the pressure above and below the surface as well as the flow of fluid at the surface. As a result of the time needed to manage individual wells in an oil field, production in large oil fields is managed by periodically (e.g., every few months) measuring fluids at collection points tied to multiple wells in an oil field and then allocating the measurements from the collection points back to the individual wells. Data collected from the periodic measurements is analyzed and used to make production decisions including optimizing future production. The collected data, however, may be several months old when it is analyzed and thus is not useful in real time management decisions. In addition to the aforementioned time constraints, multiple analysis tools may be utilized which making it difficult to construct a consistent analysis of a large field. These tools may be multiple physics-based simulators or analytical equations representing oil, gas, and water flow and processing.
In order to improve efficiency in oil field management, sensors have been installed in oil fields in recent years for continuously monitoring temperatures, fluid rates, and pressures. As a result, production engineers have much more data to analyze than was generated from previous periodic measurement methods. However, the increased data makes it difficult for production engineers to react to the data in time to respond to detected issues and make real time production decisions. For example, current methods enable the real time detection of excess water in the fluids produced by a well but do not enable an engineer to quickly respond to this data in order to change valve settings to reduce the amount of water upon detection of the excess water. Further developments in recent years have resulted in the use of computer models for optimizing oil field management and production. In particular, software models have been developed for reservoirs, wells, and gathering system performance in order to manage and optimize production. Typical models used include reservoir simulation, well nodal analysis, and network simulation physics-based or physical models. Currently, the use of physics-based models in managing production is problematic due to the length of time the models take to execute. Moreover, physics-based models must be “tuned” to field-measured production data (pressures, flow rates, temperatures, etc.) for optimizing production. Tuning is accomplished through a process of “history matching,” which is complex, time consuming, and often does not result in producing unique models. For example, the history matching process may take many months for a specialist reservoir or production engineer. Furthermore, current history match algorithms and workflows for assisted or automated history matching are complex and cumbersome. In particular, in order to account for the many possible parameters in a reservoir system that could effect production predictions, many runs of one or more physics-based simulators would need to be executed, which is not practical in the industry.
It is with respect to these and other considerations that the present invention has been made.